Missing Aggregate Dynamics:

Transcription

1 Discussion of Missing Aggregate Dynamics: On the Slow Convergence of Lumpy Adjustment Models by D. Berger, R. Caballero and E. Engel Marc Giannoni Federal Reserve Bank of New York Workshop on Price Dynamics, Central Bank of Chile Santiago, August 6, 2014 The views are solely those of the discussant and do not necessarily reflect those of the FRB of New York or the Federal Reserve System.

2 Impressive Paper 1. Addresses important questions: How rapidly are prices adjusting? How persistent are price changes? Key to understand source of business cycle fluctuations, effects of monetary policy 2. Proposes a reconciliation between aggregate data and micro data: Aggregate evidence: Price indices "sticky" and aggregate inflation persistent Disaggregated prices: prices less sticky; inflation largely transient Difference in persistence between aggregate and sectoral inflation persistence due to estimation bias 3. Theory: Proposes a theory (formula) for the persistence bias Shows that the downward bias can be important Bias disappears very slowly Provides a solution to address the bias Illustrates problem and solution in two relevant examples

3 Calvo and the Missing Persistence In Calvo model with continuum of agents/products, the aggregate price index in a sector evolves according to where λ s = freq. of price changes, p st * = opt. reset price Inflation follows Persistence of inflation in sector s: ρ s = 1 - λ s

4 Calvo and the Missing Persistence BK 2004: Get unbiased estimate of λ s : Run OLS regressions on sectoral inflation Find: Two interpretations: Linear regression model overestimates freq. of price adjustment / underestimates the inflation persistence and/or Assumptions of Calvo model do not hold BCE: Calvo assumes continuum of agents in each sector (N= ) With N finite: is biased upward

5 Suppose: Main Proposition

6 Main Proposition s Implications Estimated inflation persistence is biased downward Bias is larger, the larger is, the smaller is N

7 Intuition Suppose N=1 Recall π st * is iid Cov(π t, π t-1 ) = 0: If firm does not change its price in t or in t-1: Cov(π t, π t-1 ) = 0 If firm changes price both in t and t-1: Cov(π t, π t-1 ) = Cov(π t *, π t-1 *) = 0 So estimated inflation persistence = 0, regardless of true λ s all that matters is that consecutive adjustments are uncorrelated

8 Important contribution: Main Proposition shows that estimating persistence is tricky, when underlying changes are lumpy Difference in persistence between aggregate and sectoral inflation persistence due to estimation bias in linear models

9 Comment #1 Is there Empirical Support for the Bias Story?

10 Empirical Support for the Bias Story? Possibly Yes Boivin-Giannoni-Mihov (2009): Disaggregated PCE deflators (about 200) Decompose into aggegate and sectoral components: π st = π st A + π st I persistence of sectoral component of inflation tends to be lower (relative to persistence of aggregate component), when variance of sectoral component is larger

11 Evidence consistent with bias story Boivin-Giannoni-Mihov (2009) 1,5 1,0 rhoi / rhoa 0,5 0,0-0,5-1, ,5-2,0 Var(idio) / var(aggregate)

12 But Alternative Explanation for same fact: Rational Inattention Mackowiak and Wiederholt (2009) Firms have a fixed processing capacity constraint Face sector-specific and aggregate shocks Optimally allocate attention to sources of fluctuations depending on effect on firms profits. Volatile sectoral shocks (high ) Firms pay more attention to sectoral conditions Prices respond rapidly to sectoral shocks High λ s, low persistence of sectoral component of inflation

13 Evidence less consistent with bias story Boivin-Giannoni-Mihov (2009): responses of sectoral prices to monetary shock (unexpected tightening): If bias were present, we should see higher persistence of price response for aggregate index than for average of sectoral responses But average sectoral response (blue dashed) very similar to response of aggregate index (black solid) months months

14 Comment #2 Key Maintained Assumption: Aggregate and Sectoral Shocks Have Same Persistence

15 Key Maintained Assumption in BCE v A and v s I follow processes with same persistence Only reason for : estimation bias But: good reasons to want to want to assume different persistence in price response to sectoral shocks and aggregate shocks Plausible that: sectoral shocks imply little persistence in inflation (no further predictable changes) Aggregate shocks imply persistent changes in inflation (future predictable changes)

16 1. Sectoral shocks and Aggregate Shocks are Different BGM (2009): π st = π A st + π I st ; c st = c A st + c I st Distribution of Correlations (PCE prices, PCE quantities) Aggregate components Sectoral components correl over [-1,+1] Correl <0 agg. supply and demand shocks Sectoral productivity shocks

17 2. Rational Inattention Mackowiak and Wiederholt (2009) Firms have a fixed processing capacity constraint Face sector-specific and aggregate shocks Optimally allocate attention to sources of fluctuations depending on effect on firms profits. Volatile sectoral shocks (high ) Firms pay more attention to sectoral conditions Prices respond rapidly to sectoral shocks High λ s, low persistence of sectoral component of inflation

18 3. Calvo model with heterogeneity Carvalho (2006); Carvalho et al. (201?) Calvo model with heterogeneity in price stickiness across sectors and strategic complementarities Firms that adjust prices infrequently have disproportionately large effect on decisions of other firms and on price index. Possible for firms to respond on average quickly to sectoral shocks but slowly to aggregate shocks

19 Very impressive paper Conclusion 1. Important result: Bias in sectoral inflation persistence can be large and disappears very slowly 2. Some evidence is consistent with presence of bias 3. But paper does not reject alternative stories, specifically the possibility that sectoral shocks and aggregate shocks affect persistence of inflation differently In fact, paper does not allow for such possibility

20

21